Tuesday, April 16, 2013

Debt, Growth, and the great Excel debacle of '13

Today news broke about an influential economic study being "de-bunked" on several levels, both due to some selective data decisions and also the sexier headline of a mistaken Excel code. (note, the de-bunked link above is to a more readable summary ... for economists or others in the audience that want to read the original critique, that can be found here)

The particular study in question is about a very important topic, particularly in recent policy debates:  the importance of the U.S. debt levels to the overall economy.  We have been locked into an endless cycle of debt ceiling debates, in part fueled by the conclusions of this study and so any suggestion that these findings are questionable could (and should) have major implications for how we think about debt and austerity moving forward. The original authors Reinhart and Rogoff (paper here) essentially found (or claimed to find) that countries with high levels of debt (specifically, a debt-to-GDP ratio of 90% or more) suffered from lower economic growth rates.  These findings have been widely cited by Paul Ryan and others as evidence that austerity plans to reduce the debt are vitally necessary for the health of the economy.  Such claims are now being shown to have been based on some shaky economic foundations.

Since Krugman and others have weighed in already, I should probably just direct you to better minds on this topic than me.  But, since Krugman probably won't answer questions for the greater Stout community and I will, I'll try to weigh in the best I can with a little commentary of my own:


Criticism 1)  Serious economic research is not generally done with Excel.  So this is, in itself, pretty shocking.  But, only snobby academic economists (like me, I suppose) would be surprised by this, so I'll let that part of it go.  Still, if you're a Harvard economist, I'm pretty sure they'll spring to buy a Stata license.  Very odd.  Stout - this is a good reason to buy me a new copy of Stata too, btw (maybe my department chair will read this and surprise me with some new software?).


Criticism 2) Honestly, a criticism that can be safely made of almost any macro-economic study is that data for countries and years was chosen selectively ... if you're looking at a cross country analysis of annual data, the sad truth of the world is that not very many countries have had reliable data for very many years.  Which means that the year you choose to start your analysis can often play a big role in your conclusions ... if you choose to start in a down year, everything that comes later looks positive.  And vice versa, if you choose to start in a boom year, that comparatively makes growth look less sunny in the future.  So, while it's very bad that this particular study *appears* to have cherry picked the start date, it is not entirely shocking.  The ideal way to do such a study would to test to see if the results generally held over several possible time periods, making them robust to this question.  The authors in this case didn't do so (or at least did not report doing so).  Because of the sensitivity of start- and end- dates in these types of longitudinal studies, it is incumbent on the researcher to ensure that their fundamental conclusion is not being driven by the window of dates they chose to examine.  They seem to have purposely left out the first 6 years of available data (following WWII) without justifying why, which is troubling.


Criticism 3)  I think a key takeaway from this situation comes from the fact that one of the dangers of relying too heavily on *just* data is that it is easy to jump to conclusions based on statistical errors and data mining.  My grad school advisor (who got it from his advisor) always recited the phrase "No Data without Theory, and No Theory without Data."  In other words, an abstract theory that has no connection to real world data is pretty useless and by the same token, running a bunch of statistical regressions with no theoretical underpinnings is also pretty useless.

The paper in question (the Reinhart and Rogoff paper) is essentially just an empirical paper with little theory behind it.  And this is not surprising, because it would be very difficult to theoretically explain why there would be some sort of magical threshold at the 90% point for a debt-to-GDP ratio.  Their findings were essentially that economies in countries over 90% grew more slowly (or actually contracted) compared to those under 90%, but this is a completely arbitrary threshold that was just found as a dichotomy in this particular data set.


Criticism 4)  Abstracting from the specifics of this paper and thinking more about the big picture question: what are the implications now for the way we think about our U.S. debt levels?

For those who have had a class with me, attended a Social Science Speaker Series event with me, Honors Colloquium, or just beers at the Waterfront, many of you have already heard my basic take on whether our current debt levels are something we should be worried about.

Fundamentally, my answer is that to think about any particular level of debt is an irrelevant and arbitrary thing which is born of our inability to separate the way we think about individual household debt vs. national debts.

We think that high debt levels are bad because we know that, personally, we are in trouble if we rack up debts we can't pay back (be they credit cards, student loans, or sub-prime mortgages).

Suppose you have a $50K a year job, and for the sake of simplicity suppose that you will make that much your entire career without a raise.  (Some members of the reading audience may feel at home with this particular assumption).  In this case, we all know that taking out a 30-yr mortgage to buy a $1.5 million mansion would be a bad idea ... even if there were no interest, it would take every penny we earned for 30 years just to pay back the $1.5 million.  So this level of debt-to-income is obviously a problem and we know that's not a good idea.

But what if you are 18 years old and making minimum wage?  In this case, your income is about $15,000 a year.  Now suppose you got into a top Ivy league school and could take out $450K worth of loans to put yourself through 7 years of college and law school.  This is the exact same debt-to-income ratio, but in this case it's probably completely fine to enter into that debt.  The reason is that when you finish your Harvard law school degree, your income is going to rise significantly from your minimum wage level and you might even end up making enough in just a few years to entirely pay off your debt.

Our fundamental problem in the public debate over U.S. debt levels is that we're thinking about debt like the first case, rather than the second.  Household credit card debt (or mortgage debt) is potentially  bad if it gets too high because we only have a window of 35-40 or so years of working before we retire, and thus not only do we have to pay off debts before we retire, we also need to save.  Otherwise, we'll starve in our retirement.  A country, on the other hand, never retires (hopefully) and thus is perpetually in a "young-adult" investment mode where we should be thinking about investments in growth, and not just be focused on the level of debt itself.  That's not to say that debt doesn't matter at all ... if you took out $450K of student loans and ended up back in a job close to minimum wage, then you're in trouble.  But if that investment in education allows you to improve your state in the future sufficiently above the initial investment, then it's worth it no matter how much the initial amount happens to be.

So my main point is that this 90% mythical cutoff for debt levels is theoretically groundless and is irrelevant in my mind.  It doesn't matter if the debt level is 400% if our income is going to grow 800%.  Similarly, it doesn't matter if our debt is zero if our income is going to shrink tomorrow.

The relevant criteria for debt, therefore, is whether the money that is being borrowed today will allow improvements in our state tomorrow.  If the answer is yes, then it's irrelevant how big the debt is.  If the answer is no, then it's irrelevant how small the debt is.

And the implication of this is particularly troubling when we look at what is first to be cut in austerity measures ... things like education funding, research and development support, NSF grants, infrastructure improvements - basically all the things that would improve the future growth of the economy.

The average U.S. citizen (and media member) is approaching thinking about U.S. debt all wrong for these reasons.  We are engaged in an infinitely-lived experiment with our nation, not a 35 year working lifespan and thus we are making grave and long-reaching mistakes by thinking in static terms with year-to-year budget balancing vs. dynamic long-run growth plans.

It's not just the average non-economist who is making this mistake, as today's news has pointed out.  Many economists are stuck thinking in terms of static models.  I'll include my own dad who I often argue with about the public finance aspects of human capital investment and growth in cases like this and who, along with some of his close friends and colleagues, actually taught Paul Ryan his economics (he was an econ major at Miami University where I grew up and my dad taught before retiring a few years ago).   [shameless name-dropping, I know].  I'm pretty sure I'm right on this one, but my dad's been known to prove me wrong before and he does have a book and AER article to hold over my head at the moment until I catch up to him.   He tends to de-emphasize the public goods and externality effects in markets and I tend to perhaps over-emphasize them.  But, in a long run, multi-generational model these things compound over time and so conclusions can be dramatically different as a result.

Regardless, I'm not arguing that we should feel free to rack up any any ridiculous level of national debt we feel like with no negative consequences.  I'm arguing that our decision to engage in borrowing today should be dependent on the long-run growth consequences of that borrowing and not based on some sort of arbitrary debt ceiling or debt-to-GDP ratio that is not based in theoretical foundations.  It seems the events of the last few days support me on this one.  Take that dad!  (I love you though!)


No comments:

Post a Comment